Warning--another of my long research posts:
Here's some statistical results I found about draft age and NBA production. For draft picks 8-30 in the first round and 1-15 in the second round:
18 year olds draftees have produced an average of 4.6 additional win shares throughout their career than would be expected for their draft position.
19 year olds produce 4.8 additional win shares compared to the same expectation.
20 year olds produce 5.2 additional win shares
21 year olds produce 0.3 additional win shares
22 year olds produce 1.1 fewer win shares
23 year olds produce 4.3 fewer win shares
24 year olds produce 5.5 fewer win shares
18 year olds draftees produced .008 additional win shares per 48 minutes than would be expected by their draft position
19 year olds: .018
20 year olds: .011
21 year olds: -.018
22 year olds: -.002
23 year olds: -.002
24 year olds: -.010
EDIT: The numbers in the following paragraph are wrong. I've corrected them in post #12471 (The rest of the numbers in this post should be correct.)
On average, 18 year olds draftees play 308 games equivalents (based on 48 min/game) in their careers
19 year olds: 295
20 year olds: 339
21 year olds: 456
22 year olds: 273
23 year olds: 222
24 year olds: 191
20 year olds have the largest variation in career production (based on standard deviation of expected win shares for draft position). 19 and 21 year olds follow close behind. The variation (based on standard deviation) is about 75% as large for 22 year olds.
Conclusion: Mostly confirms accepted wisdom -- if you want better career and high end production, a 19 or 20 year old is more likely to over-perform their draft position. 19 year olds give great average production, but seem to be a bit of a flameout risk.
A 21 year old will give you the most games on average, but at below average rates per 48. 22 year olds perform a bit below average across the board, but are less risky (But the standard deviations are quite high for all age groups in comparison to the average under- or over-performing rates, so you quite clearly shouldn't put all your eggs into this basket of averages based on age. It's just something that can maybe help on the margins.)
Boring details on methodology follows:
I used the 10 closest drafts (1999-2008) that would give me a fair sample (allow time for player development, since I tracked total win shares). There were 372 total players drafted in the range between first round #8 and second round #15.
Then I computed an expected win share (and win share/48) score for each draft position (basketballreference.com data). Because there's quite a bit of variation, I averaged not only on the draft position itself, but on the two picks on either side. For example, the expected win shares of draft position #21 comes from the average total win shares of picks 19, 20, 21, 22, 23.
Then I computed the average difference between expected and actual win shares (& WS48) for each age cohort in the draft. I used age as of December 31 in the draft year.
There are some issues that make this methodology an imperfect test. You can probably figure them out as well or better than I can. But I think it's a pretty good test nevertheless. I adapted the idea from an article I saw in Bleacher Report (interestingly enough):
https://bleacherreport.com/articles...rs-would-both-benefit-from-age-limit-increase
But while the methodology is similar, I reached a rather different conclusion from that original study (though I asked a somewhat different question -- I used age and included non-college players; the original used college class and only included college draftees). I think my methodology was better, though you can judge that if you want.